Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Insect. Soc. (2010) 57:73–82 DOI 10.1007/s00040-009-0052-y Insectes Sociaux RESEARCH ARTICLE Influence of driver ant swarm raids on earthworm prey densities in the Mount Kenya forest: implications for prey population dynamics and colony migrations C. Schöning • C. Csuzdi • W. Kinuthia J. O. Ogutu • Received: 30 March 2009 / Revised: 9 October 2009 / Accepted: 21 October 2009 / Published online: 8 November 2009 Ó Birkhäuser Verlag, Basel/Switzerland 2009 Abstract African driver ants are nomadic social mesopredators feeding on a highly diverse array of prey species at different trophic levels. Colonies of certain driver ant species have a biomass which can equal that of mediumsized mammalian carnivores and the ultimate cause of their nomadic life-style is thought to be local prey depletion. The impact of driver ant swarm raids is therefore expected to be strong but the degree to which they reduce prey populations has not been quantified and it is unknown whether these spectacular predators exert significant topdown effects. We examined the combined effect of driver ant (Dorylus molestus) and swarm-attending bird (Alethe poliocephala) predation on the population dynamics of earthworms, which constitute the ants’ main prey type in Electronic supplementary material The online version of this article (doi:10.1007/s00040-009-0052-y) contains supplementary material, which is available to authorized users. C. Schöning (&) Department of Biology, Centre of Social Evolution, University of Copenhagen, Universitetsparken 15, 2100 Copenhagen, Denmark e-mail: [email protected] C. Schöning Länderinstitut für Bienenkunde, Friedrich-Engels-Strasse 32, 16540 Hohen Neuendorf, Germany C. Csuzdi Systematic Zoology Research Group of Hungarian Academy of Sciences, Budapest, Hungary C. Csuzdi Department of Zoology, Hungarian Museum of Natural History, P.O. Box 137, 1431 Budapest, Hungary the montane forest of Mount Kenya. Pre-raid earthworm biomass densities in the soil layer down to a depth of 8 cm varied by a factor of 31. The immediate effect of swarm raids was a reduction in earthworm numbers in this layer, but 8 days later earthworm numbers had recovered to preraid levels. When earthworm biomass densities were compared, no significant effect of swarm raids was detected. The estimated proportion of earthworm prey biomass extracted from 0 to 8 cm layer by driver ants and birds together was about 2.2%. Although colony distribution was overdispersed as expected based on knowledge of D. molestus migratory behaviour, predation events were highly localized. Predation frequency was low (once every 62 days on average) and highly variable. These results indicate that earthworm prey is highly abundant but at the same time so difficult to harvest that swarm raids exert only a marginal influence on earthworm populations. Longerterm studies would be required to determine whether W. Kinuthia Department of Invertebrate Zoology, National Museums of Kenya, 40658, 00100 Nairobi, Kenya J. O. Ogutu International Livestock Research Institute, 30709, 00100 Nairobi, Kenya J. O. Ogutu Universität Hohenheim, Institut für Pflanzenbau und Grünland, Fruwirthstrasse 23, 70599 Stuttgart-Hohenheim, Germany 74 earthworm populations are limited by swarm raids. The small impacts of individual raids and rapid recovery of earthworm prey populations likely underlie the low frequency of migrations and short distances travelled by migrating colonies of D. molestus. Keywords Ant-following birds Army ants Dorylus molestus Nomadism Introduction Terrestrial carnivores can have strong direct effects on the structure and dynamics of prey communities via predator– prey interactions (Sih et al., 1985; Schoener and Spiller, 1996; Frank, 2008; Dunham, 2008). Such effects can cascade beyond the prey base to lower trophic levels (Schmitz, 2003; Terborgh et al., 2001). If the effects of predation are strong and density-dependent, then carnivores can limit the population growth of prey species occurring at low densities and constituting their primary food source in environments with few suitable alternative prey species (Sinclair, 2003). Driver ants [Dorylus (subgenus Anomma) spp.; Savage, 1847; see also Kronauer et al., 2007] hunt by massive swarm raids on the forest floor and up in the vegetation in which hundreds of thousands or even a few million ant workers form a dense carpet that sweeps through areas of up to 1,000 m2 or more in a single day (Leroux, 1982). Such raid swarms can be 20 m or more wide (Leroux, 1982) and attract birds feeding on animals attempting to escape from the marauding ants (Keith et al., 1992; Peters et al., 2008). Colonies of the West African species Dorylus nigricans may contain as many as 7 million adult workers and have a total fresh mass of 50 kg (Leroux, 1982) which is equivalent to that of a female leopard (Kingdon, 1997). Driver ants have an extremely wide prey spectrum, ranging from solitary insects (mostly immature stages), Annelida (earthworms), Myriapoda (millipedes and centipedes), Crustacea (crabs, woodlice), Gastropoda (slugs and snails), Arachnida (spiders, harvestmen) to large social insect colonies (ants, honey bees), occasionally even including vertebrates (snakes, frogs, mammal carcasses) (Savage, 1847; Gotwald, 1995; Schöning et al., 2008). Thus, they affect prey species at many different trophic levels and in both aboveground and belowground food webs. Several vertebrates such as chimpanzees, mongooses and pangolins as well as subterranean army ants of the subgenus Dorylus (Typhlopone) feed on driver ants (Gotwald, 1995; Kingdon, 1997), so that they are not top predators. Colonies often move to new nest sites and the ultimate cause for these migrations is thought to be local prey depletion (Wilson, 1958; Gotwald, 1995). C. Schöning et al. Although several authors have inferred that driver ants have an intense impact on or may even limit populations of their prey (e.g. Vosseler, 1905; Weber, 1943; Dejean et al., 1999), to date no study has quantified the amounts of harvested prey in relation to their availability or examined the length of time necessary for prey populations to recover to original levels following raids. Moreover, the spatiotemporal patterns of driver ant raids at the habitat level have not yet been thoroughly documented. However, data on these phenomena are essential for understanding the influence of driver ant predation on the population dynamics and diversity of aboveground and belowground invertebrate communities in African forests and for elucidating the evolution of the nomadic habits of driver ants. While the nomadic patterns of the neotropical swarmraiding ant Eciton burchellii are highly stereotypical (Franks and Fletcher, 1983), migrations of driver ant colonies occur at irregular and much lower frequencies (Gotwald, 1995; Schöning et al., 2005a). Here we analyse the predator–prey relationship between the driver ant Dorylus molestus and earthworms. Earthworms are an important prey for many driver ant species (Gotwald, 1974; Schöning, unpubl. data) and represent the ants’ main food type at our study site in the montane forest of Mount Kenya (Schöning et al., 2008). D. molestus is widely distributed throughout eastern Africa (from Ethiopia to Mozambique) and searches for prey not only in the leaf-litter, on the forest floor and up on the vegetation but also intensely in the upper soil layers. During raids groups of workers often stay behind when the advancing swarm has passed and start digging. Minutes or hours later earthworms surge out of these holes in usually unsuccessful attempts to escape (see picture in Supplementary online material) or are transported out of them in pieces. The pivotal importance of earthworms in soil biology has been well documented. In tropical rainforests they represent about 50% of the biomass of the soil macrofauna (Fragoso and Lavelle, 1992). Many studies have emphasized that earthworms are powerful regulators of soil processes, contributing to the maintenance of soil structure and the regulation of soil organic matter dynamics (Lavelle, 1997; Brussaard, 1998). They mix organic and mineral materials producing organo-mineral complexes that influence soil structure and fundamental soil processes such as carbon mineralization, nitrogen fixation, nitrification, and other processes (Hooper et al., 2000). They also produce physical structures, like burrows, that are not only essential in maintaining soil porosity but also constitute specific sites of distinct soil processes (Perreault and Whalen, 2006). Therefore, earthworms are considered to be ecosystem engineers (Lavelle et al., 1997). Exclusion experiments have shown that predation by birds and mammals can have strong effects on earthworm Impact of driver ant swarm raids on earthworms populations in afrotropical forests (Dunham, 2008). When predators keep earthworm population densities below carrying capacity, this might result in reduced leaf-litter decomposition. We compared the pre- and post-raid earthworm densities, estimated raid impact based on prey retrieval data and examined the frequency and spatial pattern of raids. The impact of individual raids on earthworm population density was small and predation frequency was rather low but highly variable. We discuss how predation by driver ants and associated birds influences earthworm population dynamics and examine implications of prey population dynamics for driver ant migratory behaviour. Study site and methods 75 Earthworm sampling plots samples taken at different times: Raid swarm Foraging trail Nest 10m Fig. 1 Scheme for sampling earthworms before, immediately after and 8 days after driver ant predation. The sampling plots are squares with a side length of 50 cm. Please note that plot locations were actually slightly off-set because of the advance of the raid swarm during the sampling of earthworm density before raids Study site The study was carried out in August 2007 and January– February 2008 at the same site in the montane forest at the eastern slope of Mount Kenya (0°140 S, 37°340 E, altitude 1,850 m a.s.l.) used by Schöning et al. (2005a, 2008). The vegetation is classified as forest dominated by Octoea usambarensis (Lauraceae), although the abundance of this tree was reduced by selective logging in the past (Bussmann 1994). Rainfall is bimodal with two rainy seasons spanning March–May and October–December (Supplementary online material). Mean annual rainfall at Chogoria Forest Station (ca. 2.2 km from the study site) was 2,178 mm during the period February 1973–March 2008. The soil at the study site has been classified as dystric and humic nitrosols (‘‘area with dense small river network: very deep, dark, red, friable clay; partly covered by shallow, dark reddish brown, humic clay’’, Speck, 1983). Voucher specimens of the D. molestus driver ants (species status: Gotwald, 1974) from this population were deposited in the Zoological Museum of the University of Copenhagen by Kronauer et al. (2006). Voucher specimens of the earthworm species found in this study have been deposited in the Oligochaeta collection of the Hungarian Natural History Museum, Budapest (Reg. No. HNHM AF/5227-5233). Impact of swarm raids on earthworm density We found swarm raids by daily monitoring of five randomly selected colonies. Once a raid swarm had been located, we observed the area around the swarm from the periphery for 15 min and noted the maximum number of attending ant-following birds and their identity (Stevenson and Fanshawe, 2002). The swarm front’s end points were then marked with small flags so that its width could later be measured using a metre-tape. At each raid swarm we sampled nine 50 9 50 cm2 plots for earthworms: the first three about 100 cm ahead of the swarm (original earthworm density), the second three plots about 100 cm behind the swarm front (the positions of these plots were at first only indicated with small flags; sampling was carried out only after all foraging activity had ceased in the particular area, this was usually the case after 1–4 h) and the last set of plots in between the other plots 8 days later. Plot locations were selected as illustrated in Fig. 1. At each plot the leaf-litter was collected and checked for earthworms before a soil monolith (down to a depth of 8 cm) was dug up with a spade and then hand-sorted for earthworms which were killed and preserved in 70% ethanol. The earthworms from the three plots per set were pooled. We chose a depth of 8 cm as previous observations had suggested that the ants do not dig deeper (see ‘‘Discussion’’). Since many individuals were incomplete (either because they had been cut by the spade used for digging or because they had autotomized), we determined the numbers of all earthworm pieces and also their total biomass (after oven-drying at 60°C for 48 h). Earthworm densities were measured in this way in August 2007 at five swarms each of five ant colonies. The reason for examining earthworm density 8 days after the raids was that army ant migration frequency depends on the recovery dynamics of prey populations (Franks 2001) and recovery of populations of solitary invertebrate species seems to take place within 1 week (Franks 1980). Only those raid areas were re-sampled 8 days later for which we could unambiguously ascertain by subsequent daily monitoring that no other swarm raid took place in the area. For comparisons, earthworm prey biomass density data were ln-transformed to better approximate a normal distribution. Prey densities before, immediately after and 8 days after raids were compared 76 C. Schöning et al. with paired t tests. The critical p value was adjusted from 0.05 to 0.025 for experimentwise error using the Bonferroni method (Sokal and Rohlf, 1995). All reported p values are for two-sided tests. To estimate the biomass range of complete individual earthworm specimens available to the ants, we additionally weighed several specimens at both size extremes. opportunistically collected earthworms captured in swarm raids of the focal colonies on days when no other data or samples were gathered from these colonies. The earthworm specimens were preserved as described above for later identification. Independent estimation of swarm raid impact based on prey composition and prey retrieval data At the start of the fieldwork in January 2008, we delimitated a study area for measuring colony density without prior knowledge of active nest locations by choosing as border points conspicuous landmarks we remembered from previous fieldwork. The northern and southern boundaries were chosen to be a river and a dirt road, respectively. Over the course of 14 days we intensively and exhaustively surveyed the area systematically for the presence of colonies without leaving any area larger than 20 9 20 m2 unvisited. All detected colonies were subsequently monitored daily and the number of colonies nesting in the study area at the end of the 14-day period was taken to represent the total number of colonies. At the end of the 14-day period all active nest locations and boundary points were recorded with a Global Positioning System (GPS) receiver (GarminÒ, Model Summit). We also prepared a map based on bearing and distance data measured by compass and metre tape in order to validate the accuracy of the first method. The map created from GPS data does not take into account the rugged topography of the study site and may thus underestimate distances and areal size. The spatial distribution of D. molestus colonies in the map based on data measured by compass and metre tape was examined by calculating the univariate L-function (Ripley, 1976; Besag, 1977) using the software PROGRAMITA (Wiegand and Moloney, 2004). Ninety-nine Monte Carlo simulations were run to test whether the observed L-values deviate from the null hypothesis of complete spatial randomness at spatial scales, r, ranging from 10 to 300 m. As an independent estimate of the impact of swarm raids on earthworm density we calculated the amounts of earthworm prey extracted from hunting areas based on prey composition and prey retrieval data. This is important because our method comparing pre- and post-raid densities does not allow distinguishing between earthworm individuals actually removed by ants and birds and those which escaped (e.g. into deeper soil layers). Prey retrieval rates were measured at exposed sections of the principal foraging trail close to the nest when the respective swarms were fully developed. Two observers counted the number of items carried towards the nest over a period of 10 min. The mean of the numbers counted by the two observers was calculated. We examined the prey composition of four prey samples (n = about 200 items each) collected from each of the same five colonies whose swarm raids were studied. Prey samples were not collected on days when the impact of swarm raids on earthworm density was examined. Items were sorted into the two categories: earthworms and nonearthworm prey. Biomass (48 h at 60°C) of earthworm prey and other prey types was determined for each of the 20 samples. The area from which prey was extracted over a 10-min period was estimated based on the mean swarm raid width (see above) and the mean swarm raid speed of 7.2 m/h (Schöning et al., 2005a). Birds attending the swarm have been observed to capture earthworms trying to escape from the ants (either within or ahead of the swarm raid area), so that the birds’ predatory activities also need to be taken into account when assessing the impact of swarm raids. The birds’ prey intake was estimated based on knowledge of feeding energetics (see Franks, 1980, p. 81). Identity of available and hunted earthworms In February 2008, we sampled earthworms at 40 randomly selected 50 9 50 cm2 plots but this time we hand-sorted not only the 0–8 cm soil layer but also the 8–30 cm layer. All the earthworm specimens collected in 2008 were first killed in 70% ethanol, then kept in 5% methanal for 4 days for fixation and subsequently stored in 70% ethanol. Later in the laboratory the specimens were identified according to Sims (1982) and their wet mass was determined. We also Colony density and distribution Spatial and temporal predation patterns From January 30th to February 23rd, 2008, we used 100 pitfall traps placed along 4 straight transects with 25 traps each (10 m distance between traps) to investigate how often a given spot in the habitat is visited by a swarm raid and how evenly predation events are spread. The transect start points and directions had been chosen before the beginning of the colony survey (see above). The pitfall traps (2.8 cm inner diameter, ca. 10 cm deep) were placed into the soil so that their upper ridges were flush with the soil surface. The traps were 2/3 filled with a mixture of 60% propyl glycol, 5– 10% ethanol and 30–35% water. Earlier experiments (four traps each placed ahead of the raid swarms of four colonies) Impact of driver ant swarm raids on earthworms showed that this method reliably recorded predation events (all of the 16 traps contained ants). Traps were checked daily for ants and the numbers of caught individuals counted. Other organisms and debris were removed and liquid was added as necessary. Results Impact of swarm raids on earthworm density The average swarm raid width of 10.60 m (±3.52 m SD, range 5.00–17.40 m, n = 25; Table 1) did not differ from the one measured by Schöning et al. (2005a) (10.30 ± 4.60 m SD; t test, t = 0.27, df = 65, p = 0.79). Swarms were attended on average by 1.4 birds (±1.12 SD, range 0– 4). All birds belonged to Alethe poliocephala (Bonaparte), a species known to attend driver ant swarms regularly (Keith et al., 1992; Peters et al., 2008). There was no significant relationship between swarm width and number of attending birds (Spearman, r = 0.13, ns, n = 25). No earthworms were found in the leaf-litter. The mean original earthworm density was 34.92 pieces (±19.13 SD, range 7–75, n = 25) or 1.34 g biomass (±1.15 SD, range 0.15–4.72, n = 25) per 0.75 m2 in the 0–8 cm soil layer. For only eight plots was it possible to determine earthworm densities 8 days later. We were unable to monitor two of the original swarm areas due to the presence of elephants; in another nine cases we noted signs of new swarm raids directly on or within a 10 m distance of the examined plots and conservatively excluded all these cases. The remaining six areas could not be re-examined due to logistical constraints or because the fieldwork ended before the 8-day period was over. The numbers of earthworm pieces in the plots ahead of raids were higher than those in plots examined directly after swarm raids (paired t test, t = 2.73, df = 24, p = 0.01), but did not differ from those measured 8 days later (paired t test, t = 1.29, df = 7, p = 0.24). Earthworm biomass in the plots ahead of raids was neither different from that in plots examined immediately after swarm raids (paired t test, t = 1.63, df = 24, p = 0.12) nor that measured 8 days later (paired t test, t = -0.53, df = 7, p = 0.61). The largest complete earthworm specimen had a biomass of 0.922 g, 307-times the biomass of the smallest specimen (0.003 g). Independent estimation of swarm raid impact based on prey composition and prey retrieval data Among the 20 prey samples examined, Annelida made up, on average, 50.08% (±22.40% SD) of prey biomass. This proportion is not significantly different from the 55.16% (±32.37% SD) obtained for the dry season months of 77 February and March in previous years at the same site by Schöning et al. (2008) (t test, t = -0.49, df = 27, p = 0.63). The mean prey item biomass was 2.02 mg (8994.74 mg/4,443 items). The observed prey retrieval rates ranged from 90 to 686 items (mean 374.75 ± 183.89 SD, n = 15) per 10 min which translates into a mean prey retrieval rate of earthworm prey biomass of 379.10 mg per 10 min. During a 10-min period, a colony retrieves prey from an area of 12.72 m2 (10.6 m swarm raid width 9 7.2 m/h swarm raid speed = 12.72 m2 per 10 min). Such an area will hold on average 22.72 g earthworm prey biomass so that the ants extract about 1.7% of the available earthworm prey. We estimated the birds’ food intake rates based on the assumption that they consume similar food amounts as neotropical ant-following birds (Franks, 1980, p. 81). If A. poliocephala birds (mass of adults about 26.55 g, Keith et al., 1992) obtain all their food at swarm raids and rely solely on earthworms for food, the average daily intake of 1.4 individuals (average number of birds present at swarm raids) would equal about 17.47 g earthworm prey biomass. If these 1.4 birds forage at swarms from dawn to dusk (13 h), they would extract an average of 0.22 g earthworm biomass per 10 min. The combined effect of driver ant and bird predation on earthworms would then amount to about 2.6% of the earthworm biomass in and above the 0–8 cm soil layer. In areas raided at night, extraction rates would be correspondingly lower. The average proportion of biomass extracted by raids over a 24 h period is then ca. 2.2%. Identity of available and hunted earthworms No earthworms were found in the leaf-litter. Among the samples of earthworms available in the 0–8 cm layer, four species were identified (Table 2): Dichogaster (Diplothecodrilus) bolaui (length of adults 25–35 mm, diameter 1.0–1.5 mm), Di. (Diplothecodrilus) affinis (length 30– 35 mm, diameter 2.0–2.5 mm, Polytoreutus huebneri (length 210–390 mm, diameter 5–7 mm), and P. annulatus (length 60–65 mm, diameter 3–4 mm). The same species except for Di. bolaui were found in the 8–30 cm layer. The proportion of Polytoreutus individuals in the two layers was different from that of Dichogaster species (Fisher’s exact test, two-tailed, p = 0.001), with Polytoreutus individuals distributed more evenly between the two layers. The two Dichogaster species are known to live in the topmost soil layer (and are categorized as epigaeic or endogaeic polyhumic; Hendrix and Bohlen, 2002). P. huebneri seems to belong to the anaecic functional type (living in permanent burrows in the soil sometimes quite deep but feeds and defecates on the surface so they regularly come up to the surface), while the soil stratum use by P. annulatus might be epi- to endogaeic. 78 Table 1 Earthworm prey density before, immediately after and 8 days after swarm raids C. Schöning et al. Swarm raid 1 A Swarm raid width (m) 8.2 Number of birds Earthworm prey density Before Immediately after Eight days later 3 38 11 24 0.991 0.241 1.267 17 2 C 14.6 0 27 1.886 2.255 3 A 12.2 2 70 50 1.089 0.730 4 D 7.0 2 56 13 1.588 0.203 5 C 5.0 0 8 11 17 0.327 0.284 0.655 12 16 14 0.519 0.533 0.838 36 6 The upper figure indicates the total number of earthworm items found in the three 50 9 50 cm2 plots, while the lower figure represents the earthworm biomass density (g per m2) Colony D 9.7 1 7 B 11.1 2 60 1.699 1.275 8 A 14.4 2 27 27 0.525 0.389 9 A 17.4 3 43 60 0.654 0.878 10 E 12.3 0 42 11 0.830 0.216 11 C 11.2 1 27 19 1.945 1.105 2.008 12 B 10.9 4 52 60 49 3.474 2.664 3.485 13 B 6.2 2 47 34 3.574 2.295 14 C 6.3 2 27 38 0.318 1.367 15 D 13.0 2 28 2 0.500 0.044 16 C 5.0 0 7 23 0.210 1.761 17 D 16.8 2 75 59 4.723 1.412 1.596 23 44 30 34 18 E 8.4 0 44 1.398 0.728 2.606 19 A 8.8 2 49 29 35 0.778 0.607 0.673 20 B 9.4 1 39 38 2.037 2.309 21 E 15.4 0 7 13 0.151 0.527 22 B 11.2 2 19 25 1.759 1.542 23 E 13.6 1 31 17 0.739 0.250 24 E 8.9 1 27 14 1.341 0.342 25 D 8.1 0 11 12 0.542 0.849 Impact of driver ant swarm raids on earthworms 79 Table 2 Identity, abundance and wet mass of available earthworms sampled in February 2008 (40 plots each measuring 50 9 50 cm2) Species Number of specimens Wet mass (g) 0–8 cm layer Dichogaster bolaui 3 0.18 Di. affinis Dichogaster spp. juv. 21 26 2.07 1.20 Polytoreutus huebneri 10 14.04 P. huebneri fragment ? juv. 0 15.39 Polytoreutus spp. fragment ? juv. 0 7.38 28 4.98 88 45.24 P. annulatus Sum 8–30 cm layer Di. affinis 3 0.22 19 32.63 P. huebneri fragment ? juv. 0 22.10 Polytoreutus spp. fragment ? juv. 0 5.10 P. huebneri P. annulatus Sum 2 0.43 24 60.48 any other afrotropical and neotropical swarm-raiding army ant species. E. burchellii densities range between 3.5 (Barro Colorado Island/Panama; Franks, 1982) and 11 colonies per 100 ha (Corcovado/Costa Rica; Swartz, 1997), whereas Leroux (1982) and Raignier and van Boven (1955) reported a density of 25 colonies per 100 ha for D. nigricans in Ivory Coast and for D. wilverthi in DR Congo, respectively. At a site in central Kenya up to four D. molestus colonies were found nesting within a 5-ha farmland area (80 colonies per 100 ha) for brief periods (Gotwald, 1995), a density that may also be reached on a similarly small-scale and for short periods at our Mount Kenya study site. The distribution of D. molestus colonies clearly deviates from complete spatial randomness at a variety of spatial scales r = 30, 40, 50, 90, 100, 140, 150, 160 and 170 m (Fig. 2). The number of neighbouring colonies found in circles of these radii centred on active nests is lower than that expected by a random spatial distribution, indicating a ‘‘repellent’’ effect of colonies on neighbours. Juv. juvenile Spatial and temporal raid patterns Among the complete individuals in the 0–8 cm layer, 47.4% had a regenerated tail, indicating a high predation pressure. In the 8–30 cm layer the proportion of individuals with a regenerated tail was lower (28.6%). Earthworm specimens with regenerated tail are easy to recognize because the regenerated tail-part is of paler colour and its segments are thinner. The two Polytoreutus species constitute about 96% of the entire earthworm biomass in the 0–30 cm layer. The earthworms in the 8–30 cm layer represented 57% of the total earthworm biomass available down to a depth of 30 cm. Among the 29 earthworms taken from D. molestus workers in swarm raids, 25 belonged to P. huebneri, three juveniles belonged to Polytoreutus but could not be identified to species level, and one specimen was identified as Di. affinis. The proportion of Polytoreutus individuals among the sample of captured earthworms was thus much higher than expected based on their relative occurrence in the 0–8 cm layer or the entire 0–30 cm layer (Fisher’s exact tests, in both cases p \ 0.01). Some traps were removed by unknown animals and predation frequencies were corrected for lost trap-days. The mean overall raid frequency was 1.61%, meaning that a given spot was visited by army ants once every 62 days. Predation events were concentrated in a few small areas (Fig. 3), all of which were in the vicinity of active nests. Raid frequencies recorded at individual traps accordingly ranged from 0 to 24%, with 80% of all traps not containing any army ants over the 25-day period. The raid frequency differed significantly between transects (Kruskal–Wallis test, H3,100 = 8.23, p = 0.042). The mean number of Colony density and distribution The area size determined based on GPS data was 62.35 ha, whereas the area value found in the other map based on bearing and distance data was 61.56 ha. Thirty-one colonies were found which translates into a colony density of 50 colonies per 100 ha (see Supplementary online material). This density is much higher than those reported for Fig. 2 Univariate L-function (Ripley) for the distribution of D. molestus colonies at study site 1 in February 2008. The upper and lower envelopes represent the fifth highest and the fifth lowest of the ranked values of L11(r) obtained from 99 Monte Carlo simulations, respectively 80 Fig. 3 Raid rates at 100 pitfall traps placed along 4 transects over the 25-day period between 30 January and 23 February 2008 D. molestus ants in the 39 traps which captured ants was 198.2 (±188.5 SD, range 3–729). Discussion Impact of D. molestus swarm raids on earthworm prey density Driver ants and the associated birds jointly reduced the number of earthworm pieces in the 0–8 cm layer significantly compared to pre-raid levels. However, no such effect was found for the earthworm biomass. The pronounced patchiness in prey density on the scale of several m2 and the huge variation in the biomass of individual earthworms probably made the detection of a significant reduction in earthworm prey biomass in our data set difficult. The variable density of attending birds at the swarm front may have further contributed to the variation in the degree by which earthworm densities are reduced. Our estimates of the impact based on prey retrieval and prey composition data moreover indicate that the effective size of the reduction in earthworm prey density was fairly small on average. Considering the large proportion of earthworm biomass in the 8–30 cm layer recorded in February 2008, we conclude that the proportion of extracted prey biomass was probably even much smaller than 2.2%. Why is the direct impact of swarm raids so small? Although the biomass of soil-dwelling earthworms is huge, the proportion of earthworms that the ants perceive to be present (i.e. the proportion of earthworms that are truly accessible) may be small because most earthworms are hidden. Moreover, hunting earthworms requires a lot of time and effort (because of the need to dig) which reduces capture rate. Immature stages of holometabolous herbivorous insects (mainly Lepidoptera and Coleoptera) are, by contrast, much easier to find, pin down and retrieve and will thus probably be harvested with higher C. Schöning et al. efficiency (prey biomass gained per unit foraging effort invested) by the opportunistic D. molestus. We make the testable prediction that the impact of swarm raids on the populations of such other prey types is stronger so that the proportion of earthworm prey in a colony’s diet is likely to increase with increasing nest residence time. Eight days after raids earthworm prey density (both in terms of numbers and biomass) was not different from preraid levels. So swarm raids do not cause a discernible reduction in local earthworm density, at least within the 8 days we examined, an outcome probably due to vertical and/or horizontal migration of earthworms. Our results suggest that D. molestus preferentially preys on the larger Polytoreutus species. Since the long-legged driver ant workers are less adept at moving through small interstices in the soil than workers of subterranean Dorylus species (Schöning et al., 2005b), the wider tunnels of large earthworms may be a welcome opportunity to find and attack earthworm prey that would otherwise be out of reach for the predominantly surface-active D. molestus. If earthworms are mainly found when ants follow their tunnels, it is conceivable that the ants sometimes hunt at greater depths than the 8 cm down to which we measured earthworm densities in the analysis of the direct impact of swarm raids. Because the earthworm biomass in the 8–30 cm layer was even greater than in the upper layer in February 2008, the proportion of earthworm biomass extracted would in any case be small. At present too little is known about the ants’ foraging behaviour in the soil, and so it is also possible that small and large earthworms are dug out of the soil even when there are no tunnels from the surface through which the ants can proceed. We collected only those individuals which came out of their tunnels trying to escape and were then captured by the ants. Larger earthworms are certainly easier to see and smaller earthworms might be less successful in surging out of the ground when attempting to escape, so that our sampling strategy may have introduced some bias. The fact that we found a significant reduction in earthworm numbers but not in biomass suggests that small species or individuals are preyed upon to a larger extent. Predation of earthworm species inhabiting different strata should in future be documented and quantified more reliably by using DNA barcoding (Hebert et al., 2003; Chang et al., 2009) to identify the earthworm pieces retrieved by ants on foraging trails to species level. Since D. molestus colonies are known to move away from their nearest neighbour when migrating to a new nest site (Schöning et al., 2005a), one may hypothesize that colony distribution should be regular and that raid rates should be homogenous in space. The first hypothesis was supported by our data, establishing migrations directed away from conspecific colonies as an alternative mechanism to produce regular colony distribution patterns in ants Impact of driver ant swarm raids on earthworms (which in more sessile species such as harvester ants often results from territorial behaviour and predation on founding queens and incipient colonies close to established colonies; Hölldobler and Wilson, 1990). The second hypothesis was not supported. The predation events we recorded over the 25-day period were concentrated in a few small areas. The observed heterogeneity of predation rates is probably scale-dependent. If the sampling scheme had been continued for several months, the predation frequency might not have differed between transects and might also have been more homogenous within transects. Although the overall impact of swarm raids is apparently small, ants can reduce earthworm prey density severely on small scales when they capture large earthworms ([0.5 g biomass, this amount exceeds the entire pre-raid biomass found in some of the 0.75 m2 plot sets). The spatial pattern of swarm raids may therefore contribute significantly to the pronounced small-scale variation in earthworm density. In spite of the apparently marginal and ephemeral reduction of earthworm densities by swarm raids and the relatively low raid frequency (on average once every 2 months) it is still possible that D. molestus and the associated bird species limit earthworm populations. Driver ant swarm raids kill and injure earthworms and influence movement patterns and may thus have a significant impact on earthworm reproduction. Indeed the proportion of earthworms with regenerated tail was very high, though it remains unclear how much other predators (such as mammals or subterranean Dorylus species) contribute to the high predation pressure. Caution is therefore warranted in interpreting the results of this short-term study. A longterm exclusion experiment would be useful to clarify whether predation by driver ants and associated birds limits earthworm populations. Ideally, the population densities of the four prey species will then be examined separately. 81 cigera (Freitas, 1994), on the one hand, and the swarmraiding myrmicine species Pheidologeton diversus which feeds also to a large extent on plant-derived food types and migrates only very rarely, on the other (Moffett, 1988). The new findings on D. molestus in the Mount Kenya forest also fit in well with these considerations. Superabundant detritivores (earthworms) represent a major fraction of the species’ diet so that it operates at a low trophic level and has a huge prey base. As a consequence colony biomass density is extraordinarily high compared to that of E. burchellii (assuming a similar colony biomass of D. molestus and E. burchellii). The estimated impact of D. molestus raids on the populations of its main prey type was small and these prey recovered quickly after raids. Colonies raiding through an area visited 8 days earlier encounter densities of this prey type that are on average as high as pre-raid densities. The costs of re-using areas recently raided are therefore much smaller than for E. burchellii colonies. And indeed, there is considerable overlap between pre- and post-migration foraging areas (Gotwald, 1995, Schöning et al., 2005a). Overall, a strategy involving low migration frequency and short migration distance is thus selectively favoured in D. molestus. Acknowledgments We are grateful to Washington Njagi and Mwenda Tiraka for help during fieldwork and the Kenya Wildlife Service and the Kenyan Ministry of Education, Science and Technology for granting research permission. We thank Klaus Riech for preparing Fig. 1. C.S. wishes to thank Jacobus Boomsma and Eduard Linsenmair for fruitful discussions on the impact of driver ant predation on prey populations. Jon Fjeldså confirmed the ID of Alethe poliocephala based on pictures, and Titus Imboma kindly provided information on ecology of this bird. Two anonymous referees made useful suggestions that helped improve the manuscript. Financial support was provided by the Alexander von Humboldt-Foundation and the Danish National Research Foundation. Prey population dynamics and D. molestus migration behaviour References The implications of prey characteristics for the evolution of optimal migration behaviour in predatory ants have been evaluated by Gotwald (1978), Franks et al. (1999) and Franks (2001). Comparing the prey spectra of the neotropical E. burchellii which feeds largely on the brood of social insects (ants and wasps) and the afrotropical D. wilverthi which feeds mostly on the immature stages of herbivorous insects, Franks (2001) argued that species preying on animals whose populations recover more slowly following a raid have to migrate both more frequently and further to avoid encountering areas from which prey has recently been harvested. This scenario is further supported by extending the comparison to include the extremely specialized and frequently migrating Leptogenys propefal- Besag J. 1977. Contribution to the discussion of Dr. Ripley’s paper. J. R. Stat. Soc. B 39: 193-195 Brussaard L. 1998. Soil fauna, guilds, functional groups and ecosystem processes. Appl. Soil Ecol. 9: 123-135 Bussmann R. 1994. The Forests of Mt Kenya – Vegetation, Ecology, Destruction and Management of a Tropical Mountain Forest Ecosystem. PhD thesis, University of Bayreuth, Germany. Chang C.-H., Rougerie R. and Chen J.-H. 2009. Identifying earthworms through DNA barcodes: pitfalls and promise. Pedobiologia 52: 171-180 Dejean A., Schatz B., Orivel J. and Beugnon G. 1999. Prey capture behavior of Psalidomyrmex procerus (Formicidae: Ponerinae), a specialist predator of earthworms (Annelida). Sociobiology 34: 545-554 Dunham A.E. 2008. Above and below ground impacts of terrestrial mammals and birds in a tropical forest. Oikos 117: 571-579 Fragoso C. and Lavelle P. 1992. Earthworm communities of tropical rain forests. Soil Biol. Biochem. 24: 1397-1408 82 Frank D.A. 2008. Evidence for top predator control of a grazing ecosystem. Oikos 117: 1718-1724 Franks N.R. 1980. The Evolutionary Ecology of the Army Ant Eciton burchelli on Barro Colorado Island, Panama. PhD thesis, University of Leeds, UK Franks N.R. 1982. A new method for censusing animal populations: the number of Eciton burchelli army ant colonies on Barro Colorado Island, Panama. Oecologia 52: 266-268 Franks N.R. 2001. Evolution of mass transit systems in ants: a tale of two societies. In: Insect Movement: Mechanisms and Consequences (Woiwod I., Reynolds D.R. and Thomas C.D., Eds). CAB International Publishing, Wallingford, UK, pp 281-298 Franks N.R. and Fletcher C.R. 1983. Spatial patterns in army ant foraging and migration. Eciton burchelli on Barro Colorado, Panama. Behav. Ecol. Sociobiol. 12: 261-270 Franks N.R., Sendova-Franks A.B., Simmons J. and Mogie M. 1999. Convergent evolution, superefficient teams and tempo in Old and New World army ants. Proc. R. Soc. London B 266: 1697-1701 Freitas A.V.L. 1994. Nest relocation and prey specialization in the ant Leptogenys propefalcigera Roger (Formicidae: Ponerinae) in an urban area in southeastern Brazil. Insect. Soc. 42: 453-456 Gotwald W.H. Jr. 1974. Predatory behavior and food preferences of driver ants in selected African habitats. Ann. Entomol. Soc. Am. 67: 877–886 Gotwald W. H. Jr. 1978. Trophic ecology and adaptation in tropical Old World ants of the subfamily Dorylinae. Biotropica 10: 161-169. Gotwald W.H. Jr 1995 Army Ants – The Biology of Social Predation. Cornell University Press. Ithaca and London. 302 pp Hebert P.D.N., Cywinska A., Ball S.L. and deWaard J.R. 2003 Biological identification through DNA barcodes. Proc. R. Soc. London B 270: 313–321 Hendrix P.F. and Bohlen P.J. 2002. Exotic earthworm invasions in North America: ecological and policy implications. BioScience 52: 1-11 Hölldobler B. and Wilson E.O. 1990. The Ants. Cambridge, Massachusetts: Belknap Press of Harvard University Press. 732 pp Hooper D.U., Bignell D.E., Brown V.K., Brussaard L., Dangerfield J.M., Wall D.H., Wardle D.A., Coleman D.C., Giller K.E., Lavelle P., van der Putten W.H., de Ruiter P.C., Rusek J., Silver W.E., Tiedje J.M. and Wolters V. 2000. Interactions between aboveground and belowground biodiversity in terrestrial ecosystems: patterns, mechanisms, and feedbacks. BioScience 50: 1049-1061 Keith S., Urban E.K. and Fry C.H. 1992. The Birds of Africa. Vol. 4. Academic Press. London. pp Kingdon J. 1997. The Kingdon Field Guide to African Mammals. Academic Press. San Diego. 450 pp Kronauer D.J.C., Schöning C. and Boomsma J.J. 2006. Male parentage in army ants. Mol. Ecol. 15: 1147-1151 Kronauer D.J.C., Schöning C., Vilhelmsen L. and Boomsma J.J. 2007. A molecular phylogeny of Dorylus army ants provides evidence for multiple evolutionary transitions in foraging niche. BMC Evol. Biol. 7: 56. http://www.biomedcentral.com/14712148/7/56 Lavelle P. 1997. Faunal activities and soil processes: adaptive strategies that determine ecosystem function. Adv. Ecol. Res. 27: 93-132 Lavelle P., Bignell D., Lepage M., Wolters V., Roger P., Ineson P., Heal O.W. and Dhillion S. 1997. Soil function in a changing world: the role of invertebrate ecosystem engineers. Eur. J. Soil Biol. 33: 159-193 Leroux J.M. 1982. Ecologie des populations de dorylines Anomma nigricans dans la Région de Lamto (Côte d’Ivoire). Publ. Lab. Zool. 22. E.N.S. Paris. C. Schöning et al. Moffett M.W. 1988. Nesting, emigrations, and colony foundation in two group-hunting Myrmicine ants (Hymenoptera: Formicidae: Pheidologeton). In: Advances in Myrmecology (Trager J.C., Ed). Leiden: E.J. Brill, pp 355-370 Perreault J.M., Whalen J.K. 2006. Earthworm burrowing in laboratory microcosms as influenced by soil temperature and moisture. Pedobiologia 50: 397-403 Peters M.K., Likare S., Kraemer M. 2008. Effects of habitat fragmentation and degradation on flocks of African ant-following birds. Ecol. Appl. 18: 847-858 Raignier A., van Boven J.K.A. 1955. Etude taxonomique, biologique et biométrique des Dorylus du sous-genre Anomma (Hymenoptera: Formicidae). Ann. Mus. R. Congo Belg. Tervuren. Sc. Zool. 2: 1-359 Ripley B.D. 1976. The second-order analysis of stationary point processes. J. Appl. Prob. 13: 255-266 Savage T.S. 1847. The driver ants of West Africa. Proc. Acad. Nat. Sci. Phil. 4: 195-200 Schoener T.W. and Spiller D.A. 1996. Devastation of prey diversity by experimentally introduced predators in the field. Nature 381: 691-694 Schöning C., Njagi W. and Franks N.R. 2005a. Temporal and spatial patterns in the emigrations of the army ant Dorylus (Anomma) molestus in the montane forest of Mt Kenya. Ecol. Entomol. 30: 532-540 Schöning C., Kinuthia W. and Franks N.R. 2005b. Evolution of allometries in the worker caste of Dorylus army ants. Oikos 110: 231-240 Schöning C., Njagi W. and Kinuthia W. 2008. Prey spectra of two swarm-raiding army ant species in East Africa. J. Zool. 274: 85-93 Schmitz O.J. 2003. Top predator control of plant biodiversity and productivity in an old-field ecosystem. Ecol. Lett. 6: 156–163 Sih A., Crowley P., McPeek M., Petranka J. and Strohmeier K. 1985. Predation, competition, and prey communities: a review of field experiments. Ann. Rev. Ecol. Syst. 16: 269–312 Sims R.W. 1982. Revision of the eastern African earthworm genus Polytoreutus (Eudrilidae: Oligochaeta). Bull. Brit. Mus. (Nat. Hist.) Zool. Ser. 43: 253-298 Sinclair A.E. 2003. Mammal population regulation, keystone processes and ecosystem dynamics. Phil. Trans. R. Soc. Lond. B 358: 1729-1740 Sokal R.S. and Rohlf F.J. 1995. Biometry. 3rd Ed. W.H. Freeman, New York. 887 pp Speck H. 1983. Mount Kenya Area. Ecological and Agricultural Significance of the Soils. Geographica Bernensia. African Studies A2 Stevenson T. and Fanshawe J. 2002. Field Guide to the Birds of East Africa. T. and A.D. Poyser. 600 pp Swartz M.B. 1997. Behavioral and Population Ecology of the Army Ant Eciton burchelli and Ant-Following Birds. PhD thesis. University of Texas, Austin, USA Terborgh J., Lopez L., Nuňez P., Rao M., Shahabuddin G., Orihuela G., Riveros M., Ascanio R., Adler G.H., Lambert T.D. and Balbas L. 2001. Ecological meltdown in predator-free forest fragments. Science 294: 1923-1926 Vosseler J. 1905. Die ostafrikanische Treiberameise (Siafu). Der Pflanzer 1: 289-302 Weber N.A. 1943. The ants of the Imatong Mountains, AngloEgyptian Sudan. Bull. Mus. Comp. Zool. Harvard Coll. 93: 263389 Wiegand T. and Moloney K.A. 2004. Rings, circles, and null-models for point pattern analysis in ecology. Oikos 104: 209-229 Wilson E.O. 1958. The beginnings of nomadic and group predatory behaviour in the ponerine ants. Evolution 12: 24-36